From 84df7fe6a2ff20663a194d3758295bee9b1436ef Mon Sep 17 00:00:00 2001 From: Serhiy Storchaka Date: Fri, 7 Nov 2014 21:43:57 +0200 Subject: Issues #814253, #9179: Group references and conditional group references now work in lookbehind assertions in regular expressions. --- Lib/re.py | 5 +++-- Lib/sre_parse.py | 33 ++++++++++++++++++++++++--------- Lib/test/test_re.py | 38 +++++++++++++++++++++++++++++++++++++- Misc/NEWS | 3 +++ 4 files changed, 67 insertions(+), 12 deletions(-) diff --git a/Lib/re.py b/Lib/re.py index 199afee..82244db 100644 --- a/Lib/re.py +++ b/Lib/re.py @@ -352,10 +352,11 @@ class Scanner: s = sre_parse.Pattern() s.flags = flags for phrase, action in lexicon: + gid = s.opengroup() p.append(sre_parse.SubPattern(s, [ - (SUBPATTERN, (len(p)+1, sre_parse.parse(phrase, flags))), + (SUBPATTERN, (gid, sre_parse.parse(phrase, flags))), ])) - s.groups = len(p)+1 + s.closegroup(gid, p[-1]) p = sre_parse.SubPattern(s, [(BRANCH, (None, p))]) self.scanner = sre_compile.compile(p) def scan(self, string): diff --git a/Lib/sre_parse.py b/Lib/sre_parse.py index b56d437..fad8014 100644 --- a/Lib/sre_parse.py +++ b/Lib/sre_parse.py @@ -66,24 +66,25 @@ class Pattern: # master pattern object. keeps track of global attributes def __init__(self): self.flags = 0 - self.open = [] - self.groups = 1 self.groupdict = {} + self.subpatterns = [None] # group 0 + @property + def groups(self): + return len(self.subpatterns) def opengroup(self, name=None): gid = self.groups - self.groups = gid + 1 + self.subpatterns.append(None) if name is not None: ogid = self.groupdict.get(name, None) if ogid is not None: raise error("redefinition of group name %s as group %d; " "was group %d" % (repr(name), gid, ogid)) self.groupdict[name] = gid - self.open.append(gid) return gid - def closegroup(self, gid): - self.open.remove(gid) + def closegroup(self, gid, p): + self.subpatterns[gid] = p def checkgroup(self, gid): - return gid < self.groups and gid not in self.open + return gid < self.groups and self.subpatterns[gid] is not None class SubPattern: # a subpattern, in intermediate form @@ -181,7 +182,21 @@ class SubPattern: elif op in UNITCODES: lo = lo + 1 hi = hi + 1 - elif op == SUCCESS: + elif op is GROUPREF: + i, j = self.pattern.subpatterns[av].getwidth() + lo = lo + i + hi = hi + j + elif op is GROUPREF_EXISTS: + i, j = av[1].getwidth() + if av[2] is not None: + l, h = av[2].getwidth() + i = min(i, l) + j = max(j, h) + else: + i = 0 + lo = lo + i + hi = hi + j + elif op is SUCCESS: break self.width = min(lo, MAXREPEAT - 1), min(hi, MAXREPEAT) return self.width @@ -709,7 +724,7 @@ def _parse(source, state): if not sourcematch(")"): raise error("unbalanced parenthesis") if group is not None: - state.closegroup(group) + state.closegroup(group, p) subpatternappend((SUBPATTERN, (group, p))) else: while 1: diff --git a/Lib/test/test_re.py b/Lib/test/test_re.py index 3fd6ab0..0fb47a8 100644 --- a/Lib/test/test_re.py +++ b/Lib/test/test_re.py @@ -557,7 +557,7 @@ class ReTests(unittest.TestCase): self.assertEqual(re.match("a.*b", "a\n\nb", re.DOTALL).group(0), "a\n\nb") - def test_non_consuming(self): + def test_lookahead(self): self.assertEqual(re.match("(a(?=\s[^a]))", "a b").group(1), "a") self.assertEqual(re.match("(a(?=\s[^a]*))", "a b").group(1), "a") self.assertEqual(re.match("(a(?=\s[abc]))", "a b").group(1), "a") @@ -571,6 +571,42 @@ class ReTests(unittest.TestCase): self.assertEqual(re.match(r"(a)(?!\s\1)", "a b").group(1), "a") self.assertEqual(re.match(r"(a)(?!\s(abc|a))", "a b").group(1), "a") + # Group reference. + self.assertTrue(re.match(r'(a)b(?=\1)a', 'aba')) + self.assertIsNone(re.match(r'(a)b(?=\1)c', 'abac')) + # Conditional group reference. + self.assertTrue(re.match('(?:(a)|(x))b(?=(?(2)x|c))c', 'abc')) + self.assertIsNone(re.match('(?:(a)|(x))b(?=(?(2)c|x))c', 'abc')) + self.assertTrue(re.match('(?:(a)|(x))b(?=(?(2)x|c))c', 'abc')) + self.assertIsNone(re.match('(?:(a)|(x))b(?=(?(1)b|x))c', 'abc')) + self.assertTrue(re.match('(?:(a)|(x))b(?=(?(1)c|x))c', 'abc')) + # Group used before defined. + self.assertTrue(re.match('(a)b(?=(?(2)x|c))(c)', 'abc')) + self.assertIsNone(re.match('(a)b(?=(?(2)b|x))(c)', 'abc')) + self.assertTrue(re.match('(a)b(?=(?(1)c|x))(c)', 'abc')) + + def test_lookbehind(self): + self.assertTrue(re.match('ab(?<=b)c', 'abc')) + self.assertIsNone(re.match('ab(?<=c)c', 'abc')) + self.assertIsNone(re.match('ab(?